Radical Termination via β-Scission Enables Photoenzymatic Allylic Alkylation Using “Ene”-Reductases

Abstract

Allylations are practical transformations that forge C–C bonds while introducing an alkene for further chemical manipulations. Here, we report a photoenzymatic allylation of α-chloroamides with allyl silanes using flavin-dependent ‘ene’-reductases (EREDs). An engineered ERED can catalyze annulative allylic alkylation to prepare 5, 6, and 7-membered lactams with high levels of enantioselectivity. Ultrafast transient absorption spectroscopy indicates that radical termination occurs via β-scission of the silyl group to afford a silyl radical, a distinct mechanism by comparison to traditional radical allylations involving allyl silanes. Moreover, this represents an alternative strategy for radical termination using EREDs. This mechanism was applied to intermolecular couplings involving allyl sulfones and silyl enol ethers. Overall, this method highlights the opportunity for EREDs to catalyze radical termination strategies beyond hydrogen atom transfer.

Graphical Abstract

Asymmetric Csp³–Csp³ bond formation is indispensable for constructing societally essential molecules.¹ Consequently, numerous catalytic methods have been developed to facilitate their construction.² Among these, reactions involving open-shell radical intermediates are desirable because they have low activation barriers and can form sterically congested centers.³ However, strategies for rendering these transformations asymmetric remain underdeveloped compared to reactions involving other types of reactive intermediates.⁴

Enzymes are ideal catalysts for asymmetric synthesis because they can precisely orient reactive intermediates, and their activity can be optimized using directed evolution.⁵ However, biocatalysts are often restricted to their natural reaction mechanisms, limiting their ability to address selectivity challenges in chemical synthesis.⁶ An ongoing goal in the field has been to develop strategies to expand the synthetic capabilities of enzymes.⁷ We recently demonstrated that flavin-dependent ‘ene’-reductases (EREDs) could catalyze non-natural C–C bond-forming reactions involving radical intermediates.⁸ In nature, EREDs catalyze the reduction of activated alkenes via a hydride transfer mechanism.⁹ We found that these enzymes will template charge-transfer (CT) complexes between various alkyl halides and the reduced flavin hydroquinone (FMN_hq) cofactor. Irradiation with visible light promotes an electron from the cofactor to the substrate. Upon mesolytic cleavage of the carbon–halogen bond, an alkyl radical is formed that can react with an alkene to forge a new C–C bond with high selectivity. Thus far, our studies have focused on radical termination via hydrogen atom transfer (HAT) from flavin semiquinone (FMN_sq) (Figure 1a).¹⁰ To expand the synthetic utility of these catalysts, we sought to develop alternative radical termination mechanisms.

Figure 1.

Mechanisms of Radical Allylation

Radical allylations are attractive reactions because they form a new C–C bond while also introducing a handle for subsequent functionalizations.¹¹ The most common reagents for these reactions are allyl stannanes, silanes, and sulfones. In these reactions, the reagent largely dictates the mechanism of elimination. For allyl stannanes and sulfones, radical termination occurs via β-scission to produce stannyl or sulfinyl radicals (Figure 1b).¹² As C–Si (76 kcal/mol) bonds are stronger than C–Sn (45 kcal/mol) or C–SO₂Ar (65 kcal/mol), allyl silanes typically undergo different elimination mechanisms.¹³ In atom transfer reactions, the alkyl radical is trapped by a halide, followed by thermal elimination to form an alkene and halosilane (Figure 1b).¹⁴ Alternatively, the presence of an oxidant can enable a radical-polar crossover mechanism where the radical is oxidized to the β-silyl cation, which readily eliminates (Figure 1b).¹⁵ We hypothesized that a β-scission or polar crossover mechanism would be available to EREDs. However, it was unclear whether either of these mechanisms could be competitive with HAT (Figure 1c).

We tested the viability of the proposed reactivity in the cyclization of allyl silane 1 to afford γ-lactam 2. We found that GluER-T36A with an NADPH turnover system consisting of glucose as a terminal hydride source and glucose dehydrogenase (GDH) under visible light irradiation, afforded the desired product in 64% yield and >99:1 er, with < 5% yield of the reductive cyclization product 3 (Table 1, entry 1). A brief screening of GluER variants previously engineered in-house revealed that GluER-T36A-K317M-Y343F (GluER-G6) was the optimal enzyme (Table 1, entry 2).¹⁶ A control experiment confirmed that the cofactor turnover system is required to achieve high yields (Table 1, entry 3). Continuous light irradiation of GluER-G6 in buffer, without any turnover system present, is sufficient to generate both FMN_hq and FMN_sq in the protein active site (Supplemental Figure 9).¹⁷ The addition of cofactor turnover system favors formation of FMN_hq, the oxidation state responsible for radical initiation. A control experiment confirms that radical initiation cannot occur from ground state FMN_hq (Table 1, entry 4).¹⁸ Additionally, when GluER-G6 is photoreduced for 72 hours to generate a mixture of the FMN_sq and FMN_hq and then substrate is added after photoirradition was stopped, no product is formed (Table 1, entry 5). These results indicate that radical initiation does not occur from either the ground state FMN_sq or FMN_hq. While initial screens were run with six equivalents of glucose, we found that two equivalents provided comparable yields (Table 1, entry 6). Finally, we ran the reaction on a preparative scale using 0.75 mol % of GluER-G6 and isolated the desired lactam product in 46% yield with >99:1 er (Table 1, entry 7).

Table 1.

Reaction Optimization

Entry	Deviation from “Initial Conditions”	Yield (%)	e.r.
1	none	64	99:1
2	GluER-G6 instead of GluER-T36A	92	99:1
3	GluER-G6 and no cofactor turnover system	72	99:1
4	GluER-G6 and no light	0	n.d.
5	GluER-G6 without cofactor turnover system and photoreduction of the enzyme prior to addition of the substrate	0	n.d
6	GluER-G6 and 2 equiv. of Glucose	92	99:1
7	0.24 mmol scale at 43.6 mM using KRED P103	46	99:1

^a.Reaction conditions: 1 (10 μmol, 3.1 mg), purified GluER enzyme (1 mol%, 100 nmol), 100 mM buffer (18 mM final substrate concentration), iPrOH (10 % v/v), glucose (60 μmol), GDH-105 (10 wt%, 0.3 mg/rxn), NADP⁺ (2 mol%), 24 hours, 25 °C. Yields determined by HPLC using a calibration curve for Entries 1–6. Reported yield for Entry 7 was for isolated and purified material. See Supplemental Information for detailed experimental procedure and additional optimization studies.

With the optimized conditions in hand, we explored the scope of the transformation (Figure 2). GluER-G6 accommodates substituents at the ortho-, meta-, and para-positions of the aromatic ring (Figure 2, 4–7). Electron-rich substrates are more reactive than electron-deficient ones. However, the enantioselectivity is high in all cases. Unsubstituted alkenes are also effective for both 5-exo-trig and 6-exo-trig cyclization, affording products in high yields but with modest levels of enantioselectivity (Figure 2, 8, 9, 10). We attribute the low enantioselectivity to the lack of substituents on the alkene moiety to help orient the substrate within the protein active site. This enzyme can also catalyze 7-exo-trig cyclizations in promising yields and enantioselectivities. Beyond aromatic substituents, aliphatic substituents are also tolerated. While the methyl-substituted substrates are only modestly selective (Figure 2, 12), larger i-propyl and c-hexyl substituents provide synthetically useful levels of enantioselectivity (Figure 2, 13, and 14). Furthermore, we found that heterocycles, such as furan, were also well tolerated (Figure 2, 15).

Figure 2.

Substrate Scope

a. Reaction conditions: amide (20 μmol), purified GluER enzyme (1 mol%, 200 nmol), 100 mM buffer (18 mM final substrate concentration), iPrOH (10 % v/v), glucose (40 μmol), GDH-105 (10 wt%, 0.4–7.5 mg/rxn), NADP⁺ (2 mol%), 48 hours, 25 °C. Yields were determined by NMR analysis using trimethoxy benzene as an internal standard. See Supplemental Information for detailed experimental procedure and additional optimization studies.

While evaluating the substrate scope, we found that aromatic substrates containing electron-withdrawing substituents provided lower yields of the desired product compared to those with electron-donating substituents. For example, meta-CF₃ amide 16 afforded the allylated product 17 in only 3% yield. Upon further analysis, we found that the reductively cyclized product 18 is formed in 34% yield (Figure 3a). While performing a direct Hammett analysis might be difficult because substitution impacts substrate binding, this result suggests that more electrophilic radicals favor radical termination via hydrogen atom transfer rather than elimination.¹⁹

Figure 3.

Electronic Effects on Reaction Outcome

Next, we sought to interrogate the mechanism of radical termination. We envisioned two possibilities, i) a radical-polar crossover mechanism where the β-silyl radical is oxidized by FMN_sq, forming FMN_hq and a β-silyl cation that can eliminate to form the alkene and silanol, or ii) β-scission of the β-silyl radical producing the product and a silyl radical which can abstract a hydrogen atom from FMN_sq to produce oxidized FMN. As the final flavin oxidation state differs between these two mechanisms, they can be distinguished using transient absorption spectroscopy.

Transient absorption spectroscopy (TAS) studies were conducted in a sealed quartz cuvette containing GluER-G6 reduced with sodium dithionite and chloroamide 1. The sample was excited with a 370 nm pulse and UV-Vis probe spectrum (400 – 750nm) was taken over various pump-probe delay times (maximum delay of 1000 ns). As these spectra contain multiple species, deconvolution was performed using global analysis and compared to the results of data obtained in our previous ERED catalyzed reductive cyclizations.¹⁰ The first-time component is mesolytic cleavage occurring with a lifetime of 10 ps, mirroring what was observed in the reductive cyclizations. This is followed by the growth a broad spectral feature that decays with a lifetime of 38 ns. As we know cyclization is fast (<700 ps) for structurally similar substrates used for reductive cyclizations, we attribute this feature to the TMS radical group undergoing β-scission.¹⁶ The extended lifetime of the radical intermediate is potentially due to stabilization of the radical by the electropositive β-silyl group.²⁰ The spectrum formed after β-scission is consistent with the absorption profile of the neutral flavin semiquinone. This feature persists with a lifetime of 150 ns before decaying to the flavin quinone. This suggests that the silyl radical formed after β-scission abstracts a hydrogen atom from the neutral flavin semiquinone to form silane and flavin quinone (Figure 3b). In traditional radical chemistry, this mechanism is disfavored because of the strength of the C–Si bond. It is possible that this mechanism is available under biocatalytic conditions through the intermediacy of a silicon-ate complex.

Our previous studies indicate that this initiation event occurs via photoexcitation of an enzyme templated charge transfer complex that forms between the substrate and FMN_hq. To confirm that this mechanism remains the case for these substrates, we prepared a sample with reduced GluER-G6 (containing FMN_hq) and added the substrate. Consistent with our previous studies, we observed a new absorption band at 495 nm, suggesting the intermediacy of a CT complex (Figure 3c).

Having established that EREDs can catalyze intramolecular allylations, we explored whether they could facilitate intermolecular reactions. Using GluER-G6 under the standard reaction conditions, we found that chloroacetamide 19 could be coupled to trimethylallylsilane 20 to afford the γ,δ-unsaturated amide 23 in 61% yield (Figure 4). Beyond allyl silanes, silyl enol ether 22 is reactive and affords a 1,4-dicarbonyl product 24 in nearly quantitative yield. Finally, we hypothesized that allyl sulfones could be effective reagents for radical allylation because of their propensity to undergo β-scission elimination. When amide 19 is supplied with allylsulfone 21, the allylated product 23 is formed in 73% yield.²¹ Collectively, these examples suggest the generality of this radical termination mechanism.

Figure 4.

Intermolecular Allylation

a. Reaction conditions: 18 (20 μmol, 2.06μL), 19–21 (80 μmol) purified GluER enzyme (1 mol%, 200 nmol), 100 mM buffer (18 mM final substrate concentration), iPrOH (10 % v/v), glucose (40 μmol), GDH-105 (10 wt%, 0.4–7.5 mg/rxn), NADP⁺ (2 mol%), 48 hours, 25 °C. Yields were determined by HPLC using calibration curves. See Supplemental Information for detailed experimental procedure and additional optimization studies.

In conclusion, we have demonstrated that EREDs can catalyze asymmetric allylations using allyl silanes and allyl sulfones. Radical termination occurs via a β-scission mechanism that is competitive with hydrogen atom transfer from FMN_sq. Beyond offering a new mechanism, this substitution pattern significantly expands the lifetime of radical intermediates within the protein active site. These observations offer new insights into non-natural chemistry with EREDs and unlock synthetic opportunities of this enzyme-catalyzed platform to enable novel, selective radical-based transformations as solutions to unaddressed selectivity challenges in radical chemistry.